Make the numbers useful for the next commercial decision.

Improving account selection, repairing a handover and expanding a campaign are different decisions. Define the question first, then the outcome, account group and observation period. That keeps the report focused on work somebody can act on.

Define won work in the terms of the business: an accepted order, signed project, placement or renewal. Record the event date, value and source. Keep contracted, invoiced and collected amounts separate, because each describes a different stage of the commercial result.

Separate new customers, repeat orders and expansion. A repeat order from an established account may be valuable without showing that a prospecting campaign worked. A project business may have a long gap between legitimate purchases without the customer being lost.

Define earlier progress too. An enquiry assigned to the right person, a qualified opportunity with evidence and a buyer’s confirmed next decision are different steps. Use the pipeline review guide to specify stage evidence rather than counting every conversation as pipeline.

For each relevant account, keep the source, responsible owner, meaningful actions, buyer decisions and outcome together. Preserve dates and original references. Include the existing relationship, open service matters and offline conversations where they explain the work.

The record lets a colleague reconstruct what happened and how Maya followed up the announcement. Establishing whether the approach produced additional revenue needs a credible comparison too.

Keep unknowns visible. A buyer’s answer about how they heard of you is one piece of evidence; a tracked click is another. Neither necessarily describes every influence on the decision. Do not fill gaps retrospectively with whichever campaign was running at the time.

Suppose one group produces six wins from 60 accounts and another produces three from 60. Those observed rates are 10% and 5%, a difference of five percentage points. The arithmetic does not explain why the groups differ.

Check who entered each group, whether they had prior relationships, the effort applied, the observation period and which deals remain open. A group selected because its accounts were already close to buying is not a neutral comparison with an unworked prospect list.

Review losses and unresolved decisions as well as wins. The win-loss review adds buyer evidence about the decision; it should not turn a few conversations into a universal explanation. Keep the number of interviews and unknown outcomes beside the findings.

Attribution assigns reporting credit across recorded interactions. Google Analytics describes rule-based and data-driven methods, including models informed by experiments. The chosen method determines how a report allocates that credit.

Your own report must still state what was observed and how it was assigned. A CRM field copied from a campaign label does not, by itself, establish the effect of that campaign. Keep attribution useful for understanding recorded paths while avoiding a stronger causal claim than its evidence supports.

In a paid campaign brief, agree conversion definitions, source capture and sales follow-up before launch. If the enquiry system loses the campaign context or the sales team cannot identify accepted opportunities, no reporting label can reconstruct dependable evidence later.

Incrementality asks what happened because an intervention was added. A controlled experiment compares a treatment with an appropriate control under an agreed design. Google’s Conversion Lift explanation describes this distinction for its advertising studies; availability and suitability depend on the actual campaign and study requirements.

Decide the outcome, assignment unit and analysis before starting. In B2B work, the unit may need to be an account or territory rather than an individual contact. Two people in the same buying group can share information, so treating them as independent groups may undermine the comparison.

Kohavi and colleagues’ controlled-experiment paper explains why randomisation, sample size and statistical power matter. Use an appropriate study design and analysis for the decision; a dashboard showing two different percentages is not itself an experiment.

With few completed outcomes, small changes in counts can move a rate sharply. Keep the raw counts, observation window and open cases visible. Decide how much evidence the question needs before labelling a campaign or workflow successful or unsuccessful.

When a credible experiment is not practical, describe the evidence as observational. Combine account histories, buyer explanations and operational measures to choose a bounded next action, while stating that other explanations remain possible. Inconclusive evidence is different from evidence that an intervention had no effect.

Avoid changing several things and then crediting whichever change is easiest to tell a story about. If a service repair is clearly needed, make it; record the reason and the other changes underway so later interpretation remains honest.

Review reliability, team use and commercial progress separately. The monthly sales review shows how to find a justified correction when wins are sparse. You can assign a missing owner because the handover needs one, while leaving the effect on sales unproven.

Record what you know, what remains uncertain and the action you have chosen. Name its owner and specify the evidence to check next. Keep the measurement definitions with the implementation plan so a change to the process also prompts a review of the records it needs.

Plan can address an agreed research question and the evidence needed to answer it. A Build campaign includes operation and improvement during the engagement. Agree the account group, measurement definitions and review responsibilities for research handover, execution or shared work. Run is optional continuation after a ClarityRev Build; review during the initial campaign is already part of that campaign.